There are many video codecs allowing for a scalability of the video data stream without transcoding, i.e. without the need for a sequential performance with decoding and encoding.
An example of such scalable video data streams are data streams which are scalable in terms of, for example, temporal resolution, spatial resolution or signal-to-noise ratio by simply leaving off some of the enhancement layers of the respective scalable video data stream. However, until now there is no video codec allowing for an computationally non-complex scalability in terms of scene sectioning. In HEVC, there are, or there have been proposed, also concepts for restricting an HEVC data stream to a picture subarea, but still same are computationally complex.
Moreover, depending on the application, the picture content to be encoded into a data stream might be in a form which may not be effectively coded within the usually offered rectangular picture areas. For example, panoramic picture content may have been projected onto a two-dimensional plane, forming the picture area, in a manner so that the projection target, i.e. the footprint of the panoramic scene onto the picture area, may be non-rectangular and even non-convex. In that case, a more efficient coding of the picture/video data would be advantageous.
Further, random access points are provided in existing video data streams in a manner causing considerable bitrate peaks. In order to reduce the negative effect resulting from these bitrate peaks one could think of a reduction in the temporal granularity of the occurrence of these random access points. However, this increases the mean time duration for randomly accessing such a video data stream and accordingly it would be advantageous to have a concept at hand which solves this problem in a more efficient way.